Connect with us

AI

Google Splits AI Ops from AGI as Hassabis Becomes Alphabet Chief Scientist

Demis Hassabis becomes Alphabet chief scientist focused on AGI while Koray Kavukcuoglu runs DeepMind day-to-day and Jeff Dean launches Discovery Loop with.

Published

on

Google and Alphabet CEO Sundar Pichai moved Demis Hassabis out of day-to-day Google DeepMind operations on August 5, naming him Chair of Google DeepMind and Alphabet’s first chief scientist so he can concentrate on AGI and scientific applications. Koray Kavukcuoglu, the unit’s longtime CTO and Google’s chief AI architect, becomes senior vice president of Google DeepMind and reports directly to Pichai.

The same day, Jeff Dean and Sanjay Ghemawat left after 27 years to launch Discovery Loop, an independent public-benefit corporation focused on automating machine learning, science and engineering. Alphabet is a founding investor and Google Cloud a technology partner. Alphabet shares fell about 4 percent on the combined news.

The Dual Roles Locked In

Hassabis hands operational control to Kavukcuoglu while keeping a strategic seat. He continues leading Isomorphic Labs AI drug-discovery work and will advise on models and research. He described AGI as “close at hand” and said the shift gives him space for “strategic and global AI issues.”

Kavukcuoglu oversees Gemini model development, frontier AI research, the Gemini app and developer teams. Pichai framed the split as a way to accelerate commercial products while letting Hassabis shape the longer AGI and science horizon.

Role Before After August 5
Demis Hassabis CEO, Google DeepMind Chair, Google DeepMind; Chief Scientist, Alphabet; CEO, Isomorphic Labs
Koray Kavukcuoglu CTO, Google DeepMind; Chief AI Architect, Google SVP, Google DeepMind (reports to Pichai); retains Chief AI Architect
Jeff Dean Chief Scientist / senior AI roles Co-founder, Discovery Loop (independent PBC)
Sanjay Ghemawat Google Senior Fellow Co-founder, Discovery Loop

Pichai wrote that the company must “accelerate all this work and stay focused on the AI frontier” while treating the AGI moment as equally urgent. Readers can find the complete language in Pichai’s full employee memo on the changes.

Kavukcuoglu’s 13-Year Path Inside DeepMind

Kavukcuoglu joined DeepMind in its early days and has spent 13 years there. He founded the deep learning team and led work on foundational systems that still power products.

  • DQN (Deep Q-Network): early deep reinforcement learning breakthrough that learned to play Atari from pixels.
  • WaveNet: neural audio generation that became the voice engine behind Google Assistant for millions of users.
  • Later systems including IMPALA and scaled generative model work that fed into Gemini.

Before DeepMind he researched at NEC Labs America and earned a PhD at NYU in Yann LeCun’s group. Pichai called him one of the world’s foremost AI experts. His new mandate is explicit product velocity on Gemini while Hassabis steps back from the daily grind. That record sits on Kavukcuoglu’s DeepMind research leadership record.

I’ve been working towards AGI my whole life, and as we enter this pivotal moment, I’m stepping into a new role as Chair of Google DeepMind & Chief Scientist of Alphabet. This will allow me to focus on long-term strategy, and accelerating scientific breakthroughs, including leaning into my work at Isomorphic to help cure disease.

Hassabis posted that statement on X the day of the announcement, adding full confidence in Kavukcuoglu and the existing exec team.

Discovery Loop and the Four Who Left

Dean announced Discovery Loop as a public-benefit corporation whose mission is to automate machine learning, science and engineering to speed discoveries. He is joined by longtime collaborator Sanjay Ghemawat, DeepMind research VP Oriol Vinyals and Google Brain co-founder Quoc Le. The four have worked together for 14 to 30 years on infrastructure and models used at global scale.

Google remains a founding investor and Cloud partner and plans research collaboration on ML systems and infrastructure. Seed interest includes Radical Ventures and Khosla Ventures. The structure mirrors other AI labs that chose public-benefit form. Dean’s exit after 27 years removes a foundational figure from MapReduce, Bigtable, TensorFlow and the neural-net era inside Google, yet the investment keeps the relationship open.

On X, observers quickly noted the optics of Google writing a check to the talent walking out the door. The dual move looks less like pure loss and more like Google planting a satellite lab it can still call on.

Why the Split Lands Now

The timing follows strong commercial signals and lingering model-release pressure. Pichai pointed to earnings momentum across Search, YouTube and Cloud plus heavy demand for Gemini among developers and businesses.

Key baselines Pichai and Hassabis cited:

  • 950 million+ monthly users on the Gemini app
  • 900 million+ downloads of the Gemma open-model family
  • Flash model in high demand; Cyber model live; recent Gemini Robotics advances
  • Cloud revenue growth of 82 percent in the prior quarter, driven heavily by AI infrastructure and TPUs

That Cloud surge is the same story tracked in Alphabet Cloud’s 82 percent AI-driven surge. At the same time Gemini 4 is still in progress with no public timeline, and some earlier model drops ran late. The reorganization lets one leader chase product cadence while another owns the longer AGI and science narrative.

Hassabis has repeatedly argued that creativity and scientific application will define the next phase. His earlier remarks at Cannes on Hassabis on creativity as the AGI bottleneck line up with the new brief.

Gemini Numbers and the Product Clock

Kavukcuoglu inherits a stack already delivering at consumer and developer scale, yet the competitive clock is unforgiving. OpenAI and Anthropic continue hiring and shipping. Google’s answer is full-stack control from TPUs and Cloud through frontier models to the Gemini app and AI Overviews in Search.

Hassabis said the Gemini models “are in good hands” and noted progress on new models “including Gemini 4.” No release window was given. Flash demand and Gemma download totals give Kavukcuoglu immediate commercial proof points. The job is to turn those into faster frontier releases without losing the research edge that DeepMind was built on.

Internal culture talk from both memos stresses continuity. Hassabis will keep advising from the new London Platform 37 offices. Josh Woodward and the rest of the exec team stay in place. The message is acceleration, not reset.

What the Dual Track Means for the Race

The 2023 merger that created Google DeepMind from DeepMind and Google Brain put Hassabis in charge of a much larger, more product-facing organization. Three years later the company is splitting the load again: pure AGI and science under an Alphabet-level chief scientist, commercial and research execution under an SVP who reports straight to the CEO.

  1. 2014: Google acquires DeepMind; Hassabis remains CEO of the lab.
  2. April 2023: DeepMind merges with Google Brain to form Google DeepMind under Hassabis as CEO.
  3. August 5, 2026: Hassabis becomes Chair + Alphabet chief scientist; Kavukcuoglu becomes SVP of Google DeepMind; Dean, Ghemawat, Vinyals and Le launch Discovery Loop with Alphabet backing.

Second-order effects are already visible. Google keeps influence over the departing icons through capital and Cloud contracts. It frees its Nobel-winning AGI advocate for external and long-horizon work while putting a 13-year internal builder on the product throttle. Stock reaction priced in short-term uncertainty. Crowd reaction on X split between “brain drain is real” and “this is how you scale past the founder-ops bottleneck.”

Whether the structure closes the gap on the next model cycle will be the first public test. Hassabis gets the time he asked for. Kavukcuoglu gets the mandate. Dean gets independence with a safety net. Google gets both a cleaner org chart and a new external lab it helped create.

Frequently Asked Questions

Who is Koray Kavukcuoglu and what did he build before this role?

Kavukcuoglu is a 13-year DeepMind veteran who previously served as CTO of Google DeepMind and Google’s chief AI architect. He founded the lab’s deep learning team after a PhD under Yann LeCun at NYU and earlier work at NEC Labs; his groups delivered DQN, WaveNet (the speech system behind Google Assistant) and later scaling contributions that fed Gemini.

What exactly is Discovery Loop and who founded it?

Discovery Loop is an independent public-benefit corporation co-founded by Jeff Dean, Sanjay Ghemawat, Oriol Vinyals and Quoc Le. Its stated mission is to automate the loop of proposing, running and evaluating experiments in machine learning, science and engineering, starting with ML research automation before expanding into hardware design, drug discovery and clean energy. Alphabet is a founding investor and Google Cloud a technology partner.

Does Demis Hassabis still work on Gemini models?

He no longer runs day-to-day operations but continues to advise Kavukcuoglu, Josh Woodward and the Google DeepMind leadership on models and research. He retains the Chair title at Google DeepMind and works with Pichai on strategic AGI matters while leading Isomorphic Labs full-time.

How large is the Gemini app user base cited in the announcement?

Pichai and Hassabis both stated the Gemini app has reached more than 950 million monthly users. Separately, the Gemma family of open models has surpassed 900 million downloads. Those figures were presented as evidence of commercial momentum accompanying the leadership shift.

Is the Jeff Dean departure hostile or competitive with Google?

Both sides described it as friendly. Pichai called Dean’s 27-year run “incredible” and said Alphabet is excited to support the new company as founding investor and Cloud partner, with planned collaboration on ML systems and infrastructure research. Dean’s own announcement framed the move as turning attention to an ambitious new endeavor after decades of joint work inside Google.

Logan Pierce is a writer and web publisher with over seven years of experience covering consumer technology. He has published work on independent tech blogs and freelance bylines covering Android devices, privacy focused software, and budget gadgets. Logan founded Oton Technology to publish clear, no nonsense tech news and reviews based on real hands on testing. He has personally tested and reviewed dozens of mid range and budget Android phones, written extensively about app privacy, and built and managed multiple WordPress publications over the past decade. Logan holds a bachelor's degree in English and studied digital marketing at a certificate level.

Continue Reading
Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Trending